Wolfspeed Meets Guidance and Falls 17% (19 August 2026): How the Same Dollar Loss Became a 3x Bigger Loss Per Share

Wolfspeed reported fourth-quarter revenue of $149.6 million on 19 August — the midpoint of the range it guided three and a half months earlier — and the stock fell about 17.3% from the prior close, most of it after hours. Both facts are true, and the gap between them is not a demand story. The company's non-GAAP loss was $116.9 million against $119.8 million a year earlier, essentially unchanged in dollars. The loss per share went from $0.77 to $2.26, because the share count fell by two-thirds when the company came out of Chapter 11. This is what a restructured income statement does to every comparison built on top of it.

What actually happened

Wolfspeed filed its fourth-quarter results exhibit after the close on Wednesday 19 August. The figures below are taken from it directly, comparing the three months ended 28 June 2026 with the three months ended 29 June 2025.

Read that table from the bottom rather than the top and it describes a company that has stopped haemorrhaging cash, stopped building fabs, and has not yet found volume. Read it from the top and it looks like a collapse. Both readings come from the same eleven rows, which is the problem with this particular income statement.

The denominator did most of the work

The single most consequential number in the release is not a revenue or a margin. It is 51,799 — diluted weighted average shares, in thousands.

A year earlier that figure was 155,631. The pre-bankruptcy equity was largely extinguished when Wolfspeed emerged from a prepackaged Chapter 11 on 29 September 2025, eliminating roughly $4.6 billion of debt, and replaced with a much smaller new share count. Nothing about the operating business changed on that date. Everything about the per-share arithmetic did.

So: a non-GAAP loss of $116.9 million on 51.8 million shares is $2.26. The prior year's $119.8 million loss on 155.6 million shares was $0.77. The company lost $2.9 million less money and reported a per-share loss almost three times larger.

The same trap sits in the full-year figures, and it is worth spelling out because it will appear in year-on-year screens for the next four quarters. Fiscal 2026 is reported in two pieces: a Predecessor stub from 30 June to 29 September 2025 with revenue of $196.8 million and net income of $420.2 million, and a Successor period from 30 September 2025 to 28 June 2026 with revenue of $468.3 million and a net loss of $415.8 million. Add the revenue lines and you get $665.1 million, down 12.2% on fiscal 2025's $757.6 million — a legitimate operating comparison. Add the bottom lines and you get positive $4.4 million, which is arithmetic nonsense: the Predecessor's profit is a $563.4 million accounting gain on reorganisation, not a year of trading.

The three things the preview said to watch

This site's preview of the print argued that three lines mattered and that the loss per share was not one of them. The market traded the loss per share. Here is what the three actually said.

The product split is the clearest signal, and it went both ways. Power Products rose sequentially from $100.1 million to $106.3 million — the pull-forward and dual-sourcing effects the company blamed for the March dip did prove temporary, which is the first sequential increase in the device business in some time. Materials Products, which had been flat at almost exactly $50 million for two quarters and looked like it had found a floor, fell again to $43.3 million. It had not found a floor.

The AI data centre line grew, and grew more slowly. Approximately 20% sequentially, after roughly 50% and roughly 30%, is still compounding but the base effect is visibly fading. On the earnings call, chief financial officer Gregor van Issum described data centre revenue as "increasing about 20% from the fiscal third quarter and more than doubling from fiscal 2025 to fiscal 2026", and the company said it had stood up a dedicated data centre solutions team in Silicon Valley alongside the launch of its fifth-generation silicon carbide MOSFET.

The third line is the one that did not arrive. Underutilisation costs — charges for capacity that exists, is staffed and is depreciating but is not producing anything — were the number that would have separated a volume problem from a product-economics problem. The fourth-quarter release does not disclose one. Chief executive Robert Feurle said the company was focused on increasing factory utilisation while producing the same revenue with less capacity consumed, which is a description of the right objective rather than a measurement of progress toward it.

The margin: the accounting gift has been fully received

GAAP gross margin came in at negative 25%, against negative 27% in the March quarter and negative 13% a year earlier. In dollars, the gross loss narrowed from $40.0 million to $36.8 million on essentially flat revenue — a $3.2 million improvement, not a turn.

The more important point is that the depreciation relief the preview identified has now completely landed. Fresh-start accounting cut property and equipment from $3,916.5 million to $691.4 million in carrying value, and the company told investors that would remove roughly $30 million a quarter of depreciation and amortisation once it had worked through inventory. D&A was $31.7 million in the June quarter against $60.4 million a year earlier — the full $30 million, delivered.

That matters because it is now in the base. Every future year-on-year margin comparison starts from a cost line that has already been permanently reduced by an accounting event. Whatever improvement comes next has to be earned with wafers.

The guide that did not move

For the first quarter of fiscal 2027 Wolfspeed guided revenue of $140-160 million, with non-GAAP gross margin expected to remain negative and non-GAAP operating expenses of $62-66 million.

That is the identical revenue range it guided for the quarter just reported. A company whose device business grew sequentially, whose AI data centre line is compounding, and whose principal end market is supposedly rewiring itself, has guided the September quarter to no growth at the midpoint. The guide is the most honest sentence in the release: management is not yet willing to underwrite a sequential increase, and van Issum's stated priorities were reducing debt and cost of capital rather than volume.

Feurle said the company had "gained confidence in its path toward profitability" while noting that demand trends in automotive and industrial markets remain difficult to predict. Both halves of that sentence are load-bearing.

Where the volume is still supposed to come from

The demand thesis has not changed, and it is worth restating because the mechanism is specific and checkable — it is the reason the AI data centre line gets attention disproportionate to its size.

Nvidia has been pushing the industry from traditional 54-volt in-rack distribution to 800-volt direct current, because the physics of the older approach stops working at megawatt scale. Delivering one megawatt to a single rack at 54 volts would need up to 200 kilograms of copper busbar. Moving to 800 VDC is projected to cut copper use by about 45% and lift end-to-end efficiency by roughly 5%, with the architecture aimed at deployments from 2027.

Higher voltage is exactly where silicon carbide beats silicon. Switching losses in a converter rise with voltage, and wide-bandgap devices hold their efficiency where silicon MOSFETs and IGBTs start to give it away.

On 6 August, Wolfspeed and LITEON announced that LITEON had qualified Wolfspeed silicon carbide for its 800 VDC sidecar power and compute rack power supply platforms. Qualification is genuine, and it is still not revenue — the company's own risk factors flag the distance between design-ins, design-wins and volume shipment. Fourteen days later, the sequential growth rate in that line was 20% rather than 30%.

The balance sheet, and where the cash came from

Cash, cash equivalents and short-term investments were $1,088.6 million at 28 June 2026, against long-term debt of $931.0 million and convertible notes of $756.5 million net of discount. Van Issum noted that holders of $46 million of convertible notes exercised a voluntary conversion to equity during the quarter, which reduces debt and adds, incrementally, to the share count that just tripled the per-share loss.

Free cash flow was negative $60.9 million in the quarter, an improvement on the March quarter's negative $90.0 million, with net property and equipment spending of just $5.0 million. On the June quarter's burn rate, $1.09 billion of liquidity is several years of runway.

But the year-on-year rise in cash — from $955.4 million to $1,088.6 million — did not come from operations, and this is the line most likely to be misread.

The $733.2 million of incentive and investment-credit reimbursements is what funded the year. And that source is now largely drawn: the investment tax credit receivable, current plus long-term, fell from $758.4 million at 29 June 2025 to $200.9 million at 28 June 2026 — a decline of roughly 74%. The runway is real, but the next stretch of it has to be paid for by the operating business or by the capital markets rather than by a receivable that has mostly been collected.

What would change the picture

Nothing in this release settles the question the preview posed; it advances it by one quarter and sharpens the test.

The version that compounds requires three things to be true at once, and only one of them showed up: Power Products growing sequentially (it did, $100.1m to $106.3m), Materials stabilising (it did not, $50.1m to $43.3m), and gross margin improving while revenue rises rather than because depreciation fell (undetermined, because the depreciation step is now complete and the utilisation figure was not disclosed). The September quarter is the first clean read, since it is the first quarter in which the D&A relief is in both the current and prior sequential period.

The broader read-across is to the same theme covered in the repricing of mature-node power chips — AI infrastructure generating demand at the unglamorous end of the semiconductor stack — and to Coherent's fourth quarter, where the same build-out showed up as a boom in one segment and a contraction in the other inside a single company. Wolfspeed is that story caught earlier, with a weaker balance sheet and a harder legacy business to escape. The contrast with Analog Devices, where 59.3% of every new sales dollar dropped through to profit, is the whole spectrum of operating leverage in one industry: one company converting incremental revenue to profit at extraordinary rates, another unable to fill the capacity it already owns.

The same day the results landed, the July FOMC minutes were published — the event covered in the note on why the dollar fell 0.72% on a hawkish document — and that is where the macro attention sat. For the dollar and the other seven majors, that is the event that matters; a silicon carbide maker's utilisation rate is not one of the five factors the meter scores. Keeping those two categories apart is most of the discipline.

Educational macro context only — not investment advice.